Porting Various Operating Systems To Pandora?

[Tinnus]

Didn't I say I was excluding the obvious market attempt or keeping compatibility? Obviously I know the eeePCs and the OQO and stuff, but I still think that's only a market effort to save software. But it's just pushing the already pushed x86 architecture to insane limits.

I honestly, as a user, see little reason to have that though. A mobile x86 will be more expensive, use more battery, and for the average uses you can just get Linux-like software and compile it for whatever CPU is you're using (as Pandora, Zaurus, etc).

I'd never want one of those RMPCs (ridiculously mobile PCs) with a 800x480 resolution running Windows Vista in a 3" screen with 10pt size fonts.

But anyway, let's see what happens since you think the market's going with x86. Of course Asus/VIA and Intel are going with it since they live on x86 stuff. It's like Sony stopping to sell BD players

I'm with Exophase. Anyone wanting to bet a Pandora?

 

[Svartalf]

I haven't gotten any response back yet. It's a serious issue that needs to be cleared up one way or another so I hope they take providing a response serious. It might be useful to post the question on their forums but I will wait a while longer. I promise to provide details of what I hear back IF I hear back!

I didn't think you were evading us on the subject- just that long pauses from the vendor in response tend to be ominous, whether it's justified or not...

I honestly, as a user, see little reason to have that though. A mobile x86 will be more expensive, use more battery, and for the average uses you can just get Linux-like software and compile it for whatever CPU is you're using (as Pandora, Zaurus, etc).

Why would it be more expensive? Why would it eat more battery? The eeePC is no more expensive than the comparable N8x0, has similar capabilities, and similar battery life. The only thing the Nokia units have on the eeePC is perhaps GPS capabilities in the case of the N810, and overall size in the case of both Nokia models. I know, I've played with an eeePC and I happen to OWN an N800 that was bought just as the thing had come out.

What you think you know isn't. This isn't to say that I think that it will come to pass, per se, but that discounting it out of the gate because of past performances isn't really the way to go.

 

[Exophase]

Wanna place a wager on that one?

I might.


You're on buddy.

The thing is, all those things you've referenced are UMPCs or small laptops. Yes, Intel and Via are moving to lower power - this has been their trend for years. But it's to power these UMPCs and MIDs and whatever. PC class machines main selling point is that they're small devices that run PC applications. What I'm referring to are not such devices, but instead cell phones, PDAs, gaming handhelds, internet tablets, and so on that have so far done fine without legacy x86 software, and have in fact developed a stronger code base for ARM. Although there are rumors that the next iPhone will be using x86 it's actually doing quite fine without it. It's certainly doing fine without x86 software, that's for sure. And iPhone is basically one of the more solid examples of what I'm getting at (not that I like iPhone).

There's really no practical reason for x86 to succeed against ARM, which continues to get more and more sophisticated. In fact, Intel's most recently announced CPU, Silverthorne (or Atom, as you've posted) looks to be a less advanced CPU than ARM's previously announced Cortex-A9. Messiah does make it look like VIA is finally taking their CPU designs more seriously but we'll have to see what the power draw is like (yes, C7 has always been impressive, but that was compared to Pentium-M's of all things)

If x86 does supplant ARM then it'll be because of the political whims of major corporations (hello Microsoft..) and because Intel, who for some reason has decided to never back a non-x86 CPU again, has strong-armed it into place with smaller dies.

Oh, and you might want to rethink your battery life comparisons for eee vs. Nokia N8x0:

Nokia N8x0: 4 hours max draw at 1500mAh
eeePC: 3.5 hours max draw at 5200mAh

Pentium-M is rated for much, MUCH higher power consumption than a midrange ARM. For a more level comparison, Silverthorne is still rated much higher than Cortex-A9.

Chip; No matter what, saying there's going to be a Pandora 2 in 1-2 years that has ANY major architectural difference, much less an entirely different SoC (we're talking about a majority of the relevant hardware that goes into this thing) is insane. If the Pandora team is that uncommitted to their product then we're in trouble.

 

[Chip]

What I'm referring to are not such devices, but instead cell phones, PDAs, gaming handhelds, internet tablets, and so on that have so far done fine without legacy x86 software, and have in fact developed a stronger code base for ARM.

Both Intel and AMD have committed themselves to getting x86 into the cell phone and handheld market. Intel's Atom line is solid first step. Silverthorne is almost there, and next year's Moorestown is expected to have ARM-like power consumption. Everybody has gotten on fine with ARM processors in these applications because there wasn't much choice. Soon there will be a choice.

In fact, Intel's most recently announced CPU, Silverthorne (or Atom, as you've posted) looks to be a less advanced CPU than ARM's previously announced Cortex-A9.

Less advanced how? The architectures are so drastically different, what is your basis for comparison?
QUOTE
A 1.8 GHz Atom processor's single thread performance is equivalent to it predecessor Intel A100 but should outperform it on applications that can leverage simultaneous multithreading, SSE3, and EM64T x64 extensions.[5] It has a 0.6W to 2.5 TDP rating that can dip down to .01W to 0.1W [6] when idle but that doesn't include the power consumption of the chipset. It features a 2-issue simultaneous multithreading, 16 stage in-order pipeline with 32KB iL1 and 24KB dL1 caches, integer and floating point execution units, x86 front end, a 512KB L2 cache and a 533MT/s front-side bus. The design is manufactured in 9M 45nm High-k metal-gate CMOS and housed in a 441-ball µFCBGA package.

I'm not sure what most of that means, but it sure sounds pretty advanced. I should also point out that Silverthorne starts shipping in June, while A9 won't be out until sometime next year.

If x86 does supplant ARM then it'll be because of the political whims of major corporations (hello Microsoft..) and because Intel, who for some reason has decided to never back a non-x86 CPU again, has strong-armed it into place with smaller dies.

And there it is - the rebuttal to your own argument. Like them or not, Intel and MS are 100% behind x86 and they're 100% behind getting into every conceivable market. Like them or not, they have the money to do it. Intel's chips are in 80ish% and MS's software runs on 90ish% of the world's PCs. They can throw money at this issue all day and still buy and sell TI a dozen times over.

For a more level comparison, Silverthorne is still rated much higher than Cortex-A9.

Silverthorne is still too power hungry for a cell phone, but Moorestown (due for release around the same time as A9) is in the ballpark. Silverthorne ranges from .6W to 2.5W, and Moorestown is expected to be half that. That puts it at .3W at the low end, which compares well to A9's .25W.

So the moral of this story is that ultramobile x86 processors are not a matter of if, but when. In a year, two at the outside, there will be Pandora-sized x86 computers, probably running some form of Windows.

Chip; No matter what, saying there's going to be a Pandora 2 in 1-2 years that has ANY major architectural difference, much less an entirely different SoC (we're talking about a majority of the relevant hardware that goes into this thing) is insane. If the Pandora team is that uncommitted to their product then we're in trouble.

Nothing about my original statement was insane. I think you're selectively re-interpreting my words. Read it again:
QUOTE
An x86-powered Pandora2 would not be out of the question in a year or two. Whether it would be desirable is obviously debatable. x86 does have some serious advantages as far as compatibility with existing software. If the big players can overcome the drawbacks of the architecture (as they are obviously trying to do), then there is no reason to think we won't see a handheld Windows gaming rig in the near future.

 

[ldesnogu]

In fact, Intel's most recently announced CPU, Silverthorne (or Atom, as you've posted) looks to be a less advanced CPU than ARM's previously announced Cortex-A9.

Less advanced how? The architectures are so drastically different, what is your basis for comparison?
QUOTE
A 1.8 GHz Atom processor's single thread performance is equivalent to it predecessor Intel A100 but should outperform it on applications that can leverage simultaneous multithreading, SSE3, and EM64T x64 extensions.[5] It has a 0.6W to 2.5 TDP rating that can dip down to .01W to 0.1W [6] when idle but that doesn't include the power consumption of the chipset. It features a 2-issue simultaneous multithreading, 16 stage in-order pipeline with 32KB iL1 and 24KB dL1 caches, integer and floating point execution units, x86 front end, a 512KB L2 cache and a 533MT/s front-side bus. The design is manufactured in 9M 45nm High-k metal-gate CMOS and housed in a 441-ball µFCBGA package.

I'm not sure what most of that means, but it sure sounds pretty advanced. I should also point out that Silverthorne starts shipping in June, while A9 won't be out until sometime next year.
Given you admit you are not sure what most of that means let me try to explain some of these points:

- Atom is in-order with a very deep pipeline, this is extremely bad for performance and hence for energy consumption.
- 2.5 max TDP disqualifies the chip for cell phones (even half of that [Moorestuff] is too much)
- 64 bit doesn't matter (yet) to the UMPC / Cell phone market.

A9 is out-of-order and has shorter pipes. A9 has Neon (kind of SSE3). A9 will come on SoCs (if you don't know what it is, OMAP3530 is a SoC: CPU + almost all needed peripherals including GPU).

QUOTE
And there it is - the rebuttal to your own argument. Like them or not, Intel and MS are 100% behind x86 and they're 100% behind getting into every conceivable market. Like them or not, they have the money to do it. Intel's chips are in 80ish% and MS's software runs on 90ish% of the world's PCs. They can throw money at this issue all day and still buy and sell TI a dozen times over.

Huh? You think TI is the only company using ARM chips? You are so far from reality
Look there:
http://www.arm.com/community/company_list....e&display=1
http://www.arm.com/community/ARM_Core_Licensees.php

Do you really think Intel could buy all these companies?

Also current Intel market cap is $127B and TI is $38B.

QUOTE
So the moral of this story is that ultramobile x86 processors are not a matter of if, but when. In a year, two at the outside, there will be Pandora-sized x86 computers, probably running some form of Windows.

On that I agree. But it's not because they will be here that they will win in the end Intel has already lost big battles.

 

[Blah]

Well the thing with X86 is its just too late in the game. Intel, AMD and VIA are just barely starting to try to reshape X86 for mobile devices when ARM has been situated well in the mobile market for a long time. I don't see X86 becoming much of a ultra mobile architecture without very aggressive strategies from Intel and others. Besides, why the hell would anyone want to run pc apps on a cell phone.

 

[Xmas]

To clear up some confusion, the graphics driver for OMAP3 is provided by the chip manufacturer, TI, not by Imagination Technologies. Regarding the driver structure, you might find this picture interesting:
http://www.omapzoom.org/wiki/tiki-index.ph...oftwareOverview

It's in the SOC that they're using on the N800 at least, it's just that they can't offer 3D mode with the design decisions TI and they made. The widescreen design of the N800/N810 causes issues with it being anything other than a 3D decellerator. I know, I looked into it...

The display controller in OMAP2420 only supports up to VGA resolution, but you can still use an external display controller. The 3D hardware has no problem rendering to a much larger framebuffer in memory (which you can do with pbuffers anyway), and it does have the fillrate to do so at a reasonable framerate. Nokia E90, which uses the same SoC, is capable of showing 3D on its large screen. What issues are you referring to?

they've not got anything special that Intel doesn't have a variant thereof in the X3xxx series of GMA parts just for starters. If I thought I could find an ATOM SOC like the OMAP3 at this point with that GPU on it, I'd look at choosing it if it's power profile was right- just to avoid any more of this sillybuggers.

That's a bold claim - what is it based on? Intel having a tile-based rendering architecture doesn't mean it's anywhere similar to SGX. That would be like claiming ATI or NVidia don't have anything special the other doesn't have because they both design immediate mode renderers.

By the way, Intel chose to license PowerVR SGX (and VXD) for Atom.

QUOTE
and have to get ImgTec to support you unless you use Symbian which is the only OS they support other than Linux right at the moment.

I think you missed a few important OSes there.

 

[Exophase]

Both Intel and AMD have committed themselves to getting x86 into the cell phone and handheld market. Intel's Atom line is solid first step. Silverthorne is almost there, and next year's Moorestown is expected to have ARM-like power consumption. Everybody has gotten on fine with ARM processors in these applications because there wasn't much choice. Soon there will be a choice.

Actually, there have been at least two major alternatives to ARM in the portable embedded market, MIPS and SH. Save for a few exceptions the two have been all but wiped out because ARM is just a better choice in most regards. x86 isn't so much of a choice as it is a burden - having uniformity in architecture has (and will continue to be) a good thing for development. If x86 does one it'll stamp out ARM entirely with sheer force of will, I doubt it'll ever exist as an alternative. I'm not an x86 hater by any means but I don't see a good practical reason for why it should be used in mobiles, "if it ain't broke, don't fix it" (and certainly don't make things worse).

Less advanced how? The architectures are so drastically different, what is your basis for comparison?

Laurent covered this (in-order vs out-of-order, and because of that an inferior ISA like x86's hurts a lot more, not to mention the various other ISA tricks in ARM), but I would also like to say that I don't really trust what Intel says about what's "equivalent" to what. They're especially known to talk about theoretical improvements that will happen if developers code in just the right way, which almost never happens (not that I dislike Intel, they're okay by me).

I'm not sure what most of that means, but it sure sounds pretty advanced. I should also point out that Silverthorne starts shipping in June, while A9 won't be out until sometime next year.

Go read A9's white paper, that'll probably sound pretty advanced too. It doesn't really matter when Silverthorne starts shipping if viable SoC's aren't made around it anyway, at least not in this context. At first we'll probably just see it used in UMPCs, where it won't look good compared to the existing Pentium-M's used.

And there it is - the rebuttal to your own argument. Like them or not, Intel and MS are 100% behind x86 and they're 100% behind getting into every conceivable market. Like them or not, they have the money to do it. Intel's chips are in 80ish% and MS's software runs on 90ish% of the world's PCs. They can throw money at this issue all day and still buy and sell TI a dozen times over.

I was never arguing that x86 couldn't take over, just that there is very little technical reason why this should happen. Anyway, like Laurent said, TI is just a drop in the bucket for what's using ARM (I'd like to see a Nintendo handheld start using x86, that'd be the day)

Silverthorne is still too power hungry for a cell phone, but Moorestown (due for release around the same time as A9) is in the ballpark. Silverthorne ranges from .6W to 2.5W, and Moorestown is expected to be half that. That puts it at .3W at the low end, which compares well to A9's .25W.

The "low end" means almost nothing, especially since the figure you got for A9 isn't that at all. This seems like a forced comparison. IF Moorestown really can half that (if Intel is so far along I don't know why they don't just skip Silverthorne) then I'd be expecting it to be closer to 1W most of the time.

So the moral of this story is that ultramobile x86 processors are not a matter of if, but when. In a year, two at the outside, there will be Pandora-sized x86 computers, probably running some form of Windows.

Sure, there will be such things. But will a tiny computer really be a commodity device? Most people absolutely have no interest to do everything on a mobile that their desktop can do, because it just isn't practical. I wouldn't argue that they'd happen, I'd more argue that x86 is far from a sure thing for taking over mobile embedded in general. Intel is powerful but they can't always push what they want. Look at RDRAM in PCs.

Nothing about my original statement was insane. I think you're selectively re-interpreting my words. Read it again:

Any kind of Pandora 2 IS out of the question for a year or two away, maybe you meant Pandora-like device. Maybe someone else will do such a device, but so far few companies outside of China seem to have been interested in gaming handhelds at all, aside from the major commercial ones of course (and I don't expect either to use x86 any time soon).

What I really don't understand is why so many people seem to want x86 handhelds so badly. There isn't really a wealth of great x86-only code anymore, aside from some commercial software, and the only thing of interest there is games. I don't think these devices are ever going to get mainstream popularity by playing PC games.

 

[MWeston]

To clear up some confusion, the graphics driver for OMAP3 is provided by the chip manufacturer, TI, not by Imagination Technologies. Regarding the driver structure, you might find this picture interesting:
http://www.omapzoom.org/wiki/tiki-index.ph...oftwareOverview


This message pretty much mimics the response I got from ImgTec so there it is. I wasn't sure if I was allowed to post a link to this picture because I didn't know if the site was supposed to be live to the world, but there it is and it clearly shows the separation between open source and licensed components.

 

[Svartalf]

This message pretty much mimics the response I got from ImgTec so there it is. I wasn't sure if I was allowed to post a link to this picture because I didn't know if the site was supposed to be live to the world, but there it is and it clearly shows the separation between open source and licensed components.

If that's what's going on, the issue's less of a problem. It's more obnoxious than it is a potentially blocking issue. If this is what's going on, you'll get grumbles, etc. but we can safely bet on this coming together.

Thank you for alleviating my concerns. I still don't like the situation as it could be better, but this one's definitely workable. Now I'm very eagerly awaiting getting my hot little hands on one. Guess this means I need to start tooling up for ARM and ES 2.0.

 

[Chip]

The "low end" means almost nothing, especially since the figure you got for A9 isn't that at all. This seems like a forced comparison. IF Moorestown really can half that (if Intel is so far along I don't know why they don't just skip Silverthorne) then I'd be expecting it to be closer to 1W most of the time.

The A9 .25W rating came from here. I don't remember where I saw the .6W-2.5W ratings for Silverthorne Atom chips, but here is some more detailed info that just came out today. I was pretty close If I'm fully understanding the article, those power ratings include (probably non-3D accelerated) graphics and several other integrated system functions. The "Moorestown = Silverthorne / 2" quote came from Wikipedia, so take it as you will:
QUOTE
At present time, the Atom platform is too power hungry for the smartphone market when the chipset power consumption is factored in and it can't compete with ARM architecture. The Moorestown platform which is the successor of the Menlow Platform will be a system-on-a-chip design that uses half the power of a Silverthorne processor. Moorestown should have a chipset-included power consumption level low enough for the smartphone market.

Any kind of Pandora 2 IS out of the question for a year or two away, maybe you meant Pandora-like device. Maybe someone else will do such a device, but so far few companies outside of China seem to have been interested in gaming handhelds at all, aside from the major commercial ones of course (and I don't expect either to use x86 any time soon).

What I meant is that an x86 Pandora 2 won't be technologically out of the question in a couple years. If Craig (or anybody else) had the money and desire, they will be able to put together an x86 device in a similar form factor, on a similar scale, and for a comparable price with off-the-shelf parts. I'll be very surprised if somebody doesn't.

 

[ldesnogu]

What I meant is that an x86 Pandora 2 won't be technologically out of the question in a couple years. If Craig (or anybody else) had the money and desire, they will be able to put together an x86 device in a similar form factor, on a similar scale, and for a comparable price with off-the-shelf parts. I'll be very surprised if somebody doesn't.

From my hacker point of view, what I find great with Pandora is its chip. If it went the x86 road, I would not consider it anymore.

 

[Zarneth]

What I meant is that an x86 Pandora 2 won't be technologically out of the question in a couple years. If Craig (or anybody else) had the money and desire, they will be able to put together an x86 device in a similar form factor, on a similar scale, and for a comparable price with off-the-shelf parts. I'll be very surprised if somebody doesn't.
But then Microsoft would try and make Vista run on it. Do we really want to go down that road ;P

 

[JakeK]

But then Microsoft would try and make Vista run on it. Do we really want to go down that road ;P

Oh the humanity! It needs to be fixed first (XP ftw!)

 

[ldesnogu]

But then Microsoft would try and make Vista run on it. Do we really want to go down that road ;P

That's not how it works with MS: Craigix et al would have to pay MS to port Vista to please some brain dead fans
There would be one benefit: Pandora 2 would have 2 GB of RAM instead of 128 MB.
A very good idea indeed.

More seriously MS has just demonstrated Windows Mobile 6.1. I bet this works on ARM CPUs

 

[greendots]

React OS should be a good OS for a handheld x86 in a few years. It wouldn't require licensing fees and provide compatibility with a lot of software. Even though it will run on ARM soon, most apps are not open source and will not run on anything other then x86.

 

[randomhack]

The "low end" means almost nothing, especially since the figure you got for A9 isn't that at all. This seems like a forced comparison. IF Moorestown really can half that (if Intel is so far along I don't know why they don't just skip Silverthorne) then I'd be expecting it to be closer to 1W most of the time.

The A9 .25W rating came from here. I don't remember where I saw the .6W-2.5W ratings for Silverthorne Atom chips, but here is some more detailed info that just came out today. I was pretty close If I'm fully understanding the article, those power ratings include (probably non-3D accelerated) graphics and several other integrated system functions.


From what I understand, the TDP ratings quoted are only for the CPU and does not include the chipset and any peripherals such as wifi or video decode etc.

 

[Svartalf]

More seriously MS has just demonstrated Windows Mobile 6.1. I bet this works on ARM CPUs

Considering that it's worked on most of the previous WinCE (You're not fooling me, MS. That's like what George Carlin said about using "Shoot" as an explitive... It's WinCE rebadged for marketing reasons. ) versions were able to run on ARM...

The thing is, what would possess you to do something daft like that, run WinCE I mean...

 

[ldesnogu]

The thing is, what would possess you to do something daft like that, run WinCE I mean...

That won't happen, at least not to me

 

[Chip]

From what I understand, the TDP ratings quoted are only for the CPU and does not include the chipset and any peripherals such as wifi or video decode etc.

The .65W TDP rating for the Atom Z500 is for the CPU only, but the .25W TDP for the Cortex-A9 is also only for the CPU. The numbers for an A9-based SoC with all the trimmings would obviously be higher, just as a complete Atom-based system would be higher.

The power usage ratings for the the next gen (Lincroft) Atom chips destined for the Moorestown SoC are only speculative and based only on Intel's marketing materials. Still, if they end up being accurate, Moorestown could be a viable Platform for handhelds and smartphones.